Due to the ability to store and process large amounts of data the last years have been seen an increased interest in the estimation of high dimensional regression models. Situations with several million unknown parameters are no longer just exceptions but rather a challenge for users and statistical software.
As a statistician I'm specialized in the estimation of such high dimensional models. By using special algorithms and software I'm able to fit huge regression models with moderate computational equipment.
High dimensional data situations occur in many areas, for example, in the analysis of medical images, modelling of huge time series as well as in longitudinal studies with several thousands of participants.
Therefore, such problems can be found in nearly any discipline, from medicine through the natural sciences to economics and the social sciences.
The combination of modern MCMC algorithms and iterative methods for sparse linear systems provide the possibility to estimate large-scale regression models with only moderate computational equipment.
In essence, the method can be classified into the framework of strucutred additive regression models. These models are able to account for spatial and temporal information as well as non-linear effects.
The advantages of the implemented Bayesian approach: no correction for multiple comparisons necessary, intuitive interpretation of results and the possibility to account for prior information.
The main assumption for a successful application of this method is that the question at hand can be seen as a regression problem: one or more independent variables have an effect on one dependent variable.
It is further required that the regression coefficients can be structured in some way. However, in most cases it is my job to answer this question.
More about me, my work and potential costs can be found here.
Do you have a question? Write me a message. Just fill out the contact form and I will contact you as soon as possible.